Open Access

ARTICLE

Visual Object Detection and Tracking Using Analytical Learning Approach of Validity Level

Yong‐Hwan Lee, Hyochang Ahn, Hyo‐Beom Ahn, Sun‐Young Lee

Dept. of Digital Contents, Wonkwang University, Iksan, Jeonbuk, Korea
Far East University, Eumseong, Chungbuk, Korea
College of Engineering, Kongju National University, Cheonan, Chungnam, Korea
Dept. of Information Security Engineering, Soonchunhyang University, Asan-si, Chungnam, Korea

* Corresponding Author: Sun‐Young Lee,   

Intelligent Automation & Soft Computing 2019, 25(1), 205-215. https://doi.org/10.31209/2018.100000056

Abstract

Object tracking plays an important role in many vision applications. This paper proposes a novel and robust object detection and tracking method to localize and track a visual object in video stream. The proposed method is consisted of three modules; object detection, tracking and learning. Detection module finds and localizes all apparent objects, corrects the tracker if necessary. Tracking module follows the interest object by every frame of sequences. Learning module estimates a detecting error, and updates its value of credibility level. With a validity level where the tracking is failed on tracing the learned object, detection module finds again the desired object. The experimental results show that the proposed approach is more robust in appearance changes, viewpoint changes, and rotation of the object, compared to the traditional method. The proposed method can track the interest object accurately in various environments.

---

Keywords

---